Fraud Detection and Avoidance

Large North American insurer increases fraud avoidance, with a ROI of >10x

Client profile

Large North American Insurer

Fraud Detection


North America

Our client was the Canadian headquarters of one of North America’s largest insurance companies. This division was responsible for group health benefits claims for both individuals and group policies. Annual claims processed are in the billions of dollars.

The challenge

The client was seeking to augment its fraud investigative capabilities with an AI system. Current fraud recoveries were less than the industry stated belief that 3% to 10% of all claims had some element of fraud or abuse.

Our solution

Daisy implemented its Fraud Detection solution for the dental and drug lines of business during this trial to meet our client’s objectives and to build a business case for enhancing fraud detection and avoidance capabilities.

  • Daisy Fraud Detection: Automated AI-based fraud detection system that would identify potential fraud for both historical claims as well as in-process claims before adjudication was completed.

The first step to building an operational AI-based fraud detection capability was to assess the amount of fraud, waste and abuse that was detectable and recoverable or avoidable. Once the magnitude of the opportunity was understood, a business case could be developed to enhance existing capabilities. The client identified that Daisy’s AI platform for insurance could be a fit to achieve the project objectives and engaged Daisy in a trial.

A typical Daisy implementation requires two months to onboard the client’s data and to work with the business operators to understand the client’s existing process and capabilities in claims processing. The client already had a set of fraud flags that were used for fraud detection and had historical labels for previous fraudulent claims and individuals. Over a six month period, Daisy delivered its system which included integrating the clients flags with Daisy’s existing probabilistic rules. Predictive models using the client’s labels were autonomously deployed, fuzzy logic peer analysis of patients and providers was autonomously deployed, and social network analysis of all entities was autonomously deployed.

The client dedicated a single investigator per line of business to work with the Daisy system. The first phase of the trial was a backward-looking review of historical claims using the Daisy system. If sufficient evidence was uncovered in the historical review, a second phase of the trial would be to use the system in an operational test to determine the actual recoveries that can be achieved using the Daisy system.


The Daisy system applied to historical data delivered a business case that showed that potential annual fraud recoveries/avoidance between $10 million to $50 million across the two lines of business was possible. The business case was sufficient to deploy the system into an operational test for a further six months. During the operational test, significant fraud, waste and abuse avoidance was achieved using the Daisy system.

>$10M in Fraud Avoidance

By working with Daisy, the client achieved >$10M in fraud, waste and abuse avoidance.


~1000 Days of Investigative Effort

Female and male insurer having a conversation and looking at a desktop computer


We’re delivering incremental value to our clients

Ready to maximize your revenue?

Let's talk

Get started now